In financial settings, like loan evaluation, it is essential that the machine learning models used, are highly accurate to avoid financial stress for lenders and applicants. More complex models are needed to analyse non-linear patterns that commonly occur in financial data. Such models are more opaque, so present a greater barrier to explainability. Lack of transparency creates issues surrounding fairness and conflict with current EU AI legislation, when used to evaluate people for access to services, such as loans. (Regulation - EU - 2024/1689 - EN - EUR-Lex 2024) The issue is that the majority of existing solutions explaining the reasoning of black box models are geared towards technical users: machine learning experts or data scientists. This leaves an under researched gap for explanations aimed at a non-technical end-user, in this case loan applicants. This dissertation selects and preprocesses an appropriate dataset, then implements, trains and tests four machine learning models of varying opacity. This was carried out with the aim of ascertaining the best performing model using the metric ROC-AUC. This tests the models over all thresholds; it is commonly used in financial settings where risk appetite varies. XGBoost outperformed Support Vector Machine, Random Forest and Logistic Regression. XGBoost is the most opaque so requires specific techniques to explain.